List your company

Prompt Injection on CSQA (ASR, MDSR)

62ASRInspector
14.1626.583951.42May 8, 2026

Evaluation results

MethodLinks
InspectorArchitecture=Tree2026.056240
AgentXposedArchitecture=Star2026.0561.338.3
AgentSafeArchitecture=Star2026.056038.3
AgentXposedArchitecture=Random2026.0559.741.7
No DefenseArchitecture=Star2026.0556.343.3
ChallengerArchitecture=Star2026.055645
AgentSafeArchitecture=Chain2026.0552.351.7
No DefenseArchitecture=Chain2026.055246.7
InspectorArchitecture=Chain2026.0550.751.7
No DefenseArchitecture=Tree2026.055056.7
ChallengerArchitecture=Random2026.0549.350
AgentXposedArchitecture=Tree2026.054953.3
AgentSafeArchitecture=Tree2026.054856.7
ChallengerArchitecture=Tree2026.054656.7
AgentSafeArchitecture=Random2026.0545.355
No DefenseArchitecture=Random2026.0544.755
ChallengerArchitecture=Chain2026.0544.460
InspectorArchitecture=Star2026.0543.756.7
AgentXposedArchitecture=Chain2026.0543.355
InspectorArchitecture=Random2026.0536.363.3
GUARDIANArchitecture=Chain2026.0531.370
GUARDIANArchitecture=Random2026.053171.7
INFA-GuardArchitecture=Chain2026.0529.771.7
GUARDIANArchitecture=Tree2026.0529.370
INFA-GuardArchitecture=Random2026.0528.771.7
GUARDIANArchitecture=Star2026.0527.773.3
G-SafeguardArchitecture=Tree2026.0527.373.3
G-SafeguardArchitecture=Random2026.0527.375
G-SafeguardArchitecture=Chain2026.0526.775
INFA-GuardArchitecture=Tree2026.0525.371.7
G-SafeguardArchitecture=Star2026.0524.775
INFA-GuardArchitecture=Star2026.0524.775
PropGuardArchitecture=Star2026.052180
PropGuardArchitecture=Random2026.0518.383.3
PropGuardArchitecture=Tree2026.051883.3
PropGuardArchitecture=Chain2026.051675